Forum Discussion

Chris_Evans's avatar
11 years ago
Solved

Spoold.exe Constant Disk IO

Just seeing if the behaviour we are experiencing is normal with a Netbackup PureDisk pool on Windows 2008 R2 - Netbackup 7.6.0.1. We are seeing a high read IO on the puredisk volume. Its hard to tell what is casing this or is it normal. After a backup we can see up to and more 120MB/sec or reads to the purdisk volume (only thing on the volume). This is with no backups or duplication jobs running. Does this process throttle itself if we start duplicating images to tape and need high read performance? No AV running on this volume, and process explorer flags spoold.exe as the cultript. 

Also is there anyway to benchmark the rehydration process to tape? I have a strong suspicion that we are not pushing data to the tape drives as fast as they can handle it.

 

  • This could be MSDP self-healing and integration checking 

    http://www.symantec.com/docs/DOC6466

    Page 217 - "NetBackup performs some of the integrity checking when the storage server is idle"

7 Replies

  • It can be that internal check you are seeing, we asked Symantec what 1000 IOPS we saw at idle time under the 7.6Beta2 was and the answer was the internal check. At that time it was an on/off thing

  • Ok, thats what I ws kinda expecting to hear, but doesn't confirm if the check process backs off if I decide to throw a workload against it ie start a duplication to tape will the check process stop?

  • sorry don't know as our 7.6Beta2 setup was purely disk based and only ran few backups for testing of certain features

  • This could be MSDP self-healing and integration checking 

    http://www.symantec.com/docs/DOC6466

    Page 217 - "NetBackup performs some of the integrity checking when the storage server is idle"

  • I would check the spoold.log , storaged.log for more details of what they're doing during that time. 

    Space compaction, queue processing, garbage collection etc.

    There is a way to check the frequency of those processes running, usually this won't be changed unless there is a backlog building up or advised by support. I assume you're not running MSDP on a master server, otherwise it could easily run into situation like this, it is not recommended.

  • Bottom line whether compacting, integrity checking or self healing. It normal and expected for MSDP to use I/O when idle.